Commerce website written in pascal

[Joanna]

https://hackaday.com/2025/10/28/web-development-in-pascal/

[cdbc]

Hi
Nice catch Joanna 
Interesting read...
Regards Benny

[MarkMLl]

Yes, it's interesting: but it still doesn't cancel the fact that as far as the rest of the World is concerned Pascal is as dead as Queen Anne.

We can't try to promote the language any more: people just aren't interested. There are however two things that we usefully /can/ do:

* Quietly remind people that most of the things considered "good practice" today originated in Pascal or its immediate successors: strong type checking, modularisation and so on.

* Suggest that Lazarus and the LCL are one of the best RAD environments, even if people don't like the underlying language.

I know that some members of this community won't like that. However as one of the greybeards I suggest that FPC+Lazarus is very similar to what in the 1980s we'd have called a "4GL": except that it has a good theoretical underpinning rather than being "yet another variant of BASIC".

Our problem is that if somebody looks at FPC+Lazarus trying to solve a specific problem, whilst we might have the library routines etc. to help him, the machine-generated RTL+FCL documentation leaves much to be desired.

MarkMLl

[Joanna]

Hi Benny I can’t really claim credit, jns gave me the link in chat.
Here is another link from him https://communitymedia.video/w/h1jdbHm8xTj5VEDuCWAbBW

[dbannon]

* Quietly remind people that most of the things considered "good practice" today originated in Pascal or its immediate successors: strong type checking, modularisation and so on.

Disagree. That only emphasizes that "its history". Better to remind them that Pascal does (and always did if you must) provide the strong typing that is again becoming fashionable. And similarly modularization.

* Suggest that Lazarus and the LCL are one of the best RAD environments, even if people don't like the underlying language.

Defintly agree.

... the machine-generated RTL+FCL documentation leaves much to be desired.

Yes, if I hear "read the source" one more time ....

On a brighter note, at a social event last night, met a young fellow who writes Java for a major telco. I said "brace yourself, I write in Pascal" and his response was "yes, strong typing, thats what we need...".

Davo

[ginoo]

I wrote myself a web development kit and I'm very happy with it. It works great. This is to say that if you want, you can build your own stuff.

[Joanna]

Don’t worry mark, many programmers will soon have plenty of spare time to come learn pascal after their non pascal programming jobs are replaced by AI. 

[MarkMLl]

Don’t worry mark, many programmers will soon have plenty of spare time to come learn pascal after their non pascal programming jobs are replaced by AI. 

We live in interesting times, but since I'm old enough to remember (and had tenuous professional involvement with) the buzz around "intelligent knowledge-based systems" in the mid-1980s I'm sceptical.

To paraphrase an observation made by Sabine Hossenfelder (which might or might not be her own): the problem with current AIs is that they do not test whether output they generate exists in or at least is compatible with their training data. An obvious example of that is the several reported instances where an AI has fabricated a plausible but fictional legal reference.

The current crop of AI systems doesn't reason, it extrapolates and hopes that its output is at least superficially plausible. And even IBM's Watson, which used somewhat different techniques to reason about Jeopardy questions etc., turned out to be virtually impossible to turn into a profitable product.

And to rub it in, if current "AI" is so great, why hasn't it already developed an AGI ** methodology or at least told its keepers what additional material it needs so that it can work on the problem?

MarkMLl

** https://en.wikipedia.org/wiki/Artificial_general_intelligence

[440bx]

I'm definitely no proponent of A.I.  I think there are many problems with it, not just technical but also ethical.

As far as the technical part, I'm starting to think of A.I as something like a text editor.  If you know how to use it then "it's a good text editor" and if you don't, it's likely to be a "not so great text editor". (of course, this is typical of any application, not just text editors... it's so much easier to blame the program than oneself.)

I see what forum user @schuler gets from A.I and I have to admit that I find it impressive and, it definitely saves a lot of time but, it's quite obvious to me that I don't have his A.I skills, it is very difficult for me to get anything that I consider really saved time from some A.I thing and, that's quite likely in part because I lack skills in that area.

The A.I thing is a real PITA in more ways than one.

[marcov]

Our problem is that if somebody looks at FPC+Lazarus trying to solve a specific problem, whilst we might have the library routines etc. to help him, the machine-generated RTL+FCL documentation leaves much to be desired.

I don't understand. Only the prototypes are machine generated, the content is handwritten. (and in the case of FPC quite ok)

[dbannon]

Our problem is that if somebody looks at FPC+Lazarus trying to solve a specific problem, whilst we might have the library routines etc. to help him, the machine-generated RTL+FCL documentation leaves much to be desired.

I don't understand. Only the prototypes are machine generated, the content is handwritten. (and in the case of FPC quite ok)

Marcov, not to argue, just inform. Some rough notes about how I find the RTL and FCL documentation -

Firstly, I struggle to identify which is RTL (89 items in doc, 66 items in src), FCL (44 items in doc), Packages (154 items in src). The RTL number seems similar but in fact almost no correlation between the doc and src entry titles.

Its possible to search the documentation from a "Search the documentation" link, eg "fcl-web" appears to go off into an endless loop.

Most of the 89 entries under RTL take you to a page containing one or several sentences about the unit. Rarely any mention or links to unit contents or how to use it.

The FCL doc entries in the docs seem to be, on average, more informative but, many are missing. As  I am unsure where, in the source, the FCL exists (is it just fpc-* or all in "packages" ?) its hard to tell just what percentage is documented. But again, fc-web does not seem to be there.

So staying on my theme of fcl-web (for no other reason than its a great package) I find that googling for it takes me to the "daily" documentation, something  I have seen before but can never find when I want it. It says -

FPC is distributed with a lot of units (well over 1500), organized in packages. For most of these no real documentation is available That includes fcl-web,

On the other hand, Lazarus documentation is actually quite good, an enormous credit to Don Siders who works tirelessly away with no thanks. Thanks Don, a really great job !

EDIT : typo in quote
Davo

[dsiders]

Our problem is that if somebody looks at FPC+Lazarus trying to solve a specific problem, whilst we might have the library routines etc. to help him, the machine-generated RTL+FCL documentation leaves much to be desired.

I don't understand. Only the prototypes are machine generated, the content is handwritten. (and in the case of FPC quite ok)

Marcov, not to argue, just inform. Some rough notes about how I find the RTL and FCL documentation -

...

On the other hand, Lazarus documentation is actually quite good, an enormous credit to Don Siders who works tirelessly away with no thanks. Thanks Don, a really great job !

@dbannon

Thank you for your kind words. They are much appreciated.

To be fair the scope for LCL+LazUtils is tiny compared to RTL+FCL. Keeping up is easier (but not easy). MvC does a great job considering that docs is not (to my knowledge) his primary task.

[MarkMLl]

On the other hand, Lazarus documentation is actually quite good, an enormous credit to Don Siders who works tirelessly away with no thanks. Thanks Don, a really great job !

I agree, and also massive thanks to MvC.

The problem is the documentation of the RTL etc., and I would highlight among other things

a) Excessive reliance on numeric constants, which as a result aren't indexed. As trivial examples, in the Sockets unit fpSend() and fpRecv() both have a flags parameter declared as a cint, with no cross-reference to available constants.

b) Lack of overall indexing of helpers etc., which as a result are effectively undocumented as far as anybody not intimately familiar with the libraries' structure and available facilities is concerned.

MarkMLl

[BSaidus]

Nice job @joanna.
I wonder what is the exact GNU/Linux distrib you use.

[Joanna]

Me? I haven’t done much lately. I tried fedora before. Don’t remember the details.